Vivo Unveils AIGC Algorithm Progress

Under the wave of AIGC, cell phone corporations that target imaging are rising their funding in AI expertise.

On July thirtieth, Vivo disclosed for the primary time its analysis and improvement achievements associated to the AIGC algorithm, together with the institution of Chongqing 3D movement seize laboratory, in addition to the discharge of a self-developed picture chip V3 with multi-concurrent AI notion. It additionally introduced that it’ll apply new-generation AI expertise in its next-generation merchandise.

Interface News reporters have discovered that in its 3D movement seize laboratory, Vivo has constructed a film-grade mild area system. This system can reconstruct “pore-level” human physique 3D fashions, supporting the analysis and iterative work of varied portrait 3D notion algorithms in Vivo. The focus is on creating core portrait 3D notion algorithm capabilities for faces, our bodies, and gestures.

The laboratory is situated within the Vivo Chongqing Intelligent Manufacturing Center. This middle can also be a key a part of Vivo’s international clever manufacturing, with its structure began in 2014, accomplished in 2017, and formally put into operation.

According to the introduction, Vivo collects human 3D knowledge and human reflectance knowledge on this laboratory. AI then learns algorithms, that are subsequently utilized to resolve some imaging issues that present {hardware} struggles with. These embody background blur results for portrait images on the hair strand degree; improvement of adaptive lighting algorithms based mostly on background model; expression modifying, coiffure modifying, age modifying based mostly on facial depth and texture data; AI-based digital dressing, AR dance, and extra utilizing human 3D knowledge and depth data.

SEE ALSO: vivo X Fold2 and X Flip Debuts in China

“3D imaging and AIGC are the core directions of Vivo’s future exploration. Currently, there are more than 100 people, most of whom have a doctoral degree,” mentioned Yu Meng, Vice President of Imaging at Vivo throughout an interview.

With the popularization of AI, cell phone corporations are accelerating their technological investments one after one other. In addition to Vivo, Honor proposed the idea of edge-side AI giant fashions in July, and in August, it’s rumored that Huawei may also incorporate the presently widespread AI giant mannequin expertise into HarmonyOS 4.

In cellular purposes, imagery has turn out to be the perfect testing floor on this technological area. Overseas, each Midjourney and OpenAI have made breakthrough progress in purposes similar to drawing and picture processing. Meanwhile, in China, a current AI digital camera app referred to as ‘Miaoya Camera’ has shortly attracted numerous customers with its easy utilization technique, inexpensive worth, and high-quality generated results.

Li Zhuo, the Senior Director of Vivo Imaging Products, believes that cellular imaging expertise is presently advancing in three areas concurrently: optics, computing energy, and algorithms. However, in comparison with the opposite two areas, optics face better challenges as it’s analog, and breaking via sure bottlenecks may be extraordinarily troublesome. It could require breakthroughs in supplies science and elementary physics. But as soon as these breakthroughs are achieved, they are going to result in an entire new world. Therefore, developments and breakthroughs in computing energy and algorithms are very obvious at this stage.

According to the on-site demonstration, counting on AI expertise, the telephone can obtain exact seize and driving of facial expressions, physique postures, and gesture interactions. It may take pictures that mix digital and actual scenes in a extra pure and harmonious means, even attaining lifelike results. Technologies similar to texture reconstruction and lighting estimation algorithms permit customers to re-edit points similar to lighting, age, coiffure, and expression based mostly on their preferences.

“In the future, algorithms will lead the way, but it doesn’t mean they run on their own. In the future, algorithms will empower traditional optics. This is the logical relationship within them,” mentioned Yu Meng.

Sign up immediately for five free articles month-to-month!